KR102553692B1 - Deep learning system for disaster detection in the underground power facility - Google Patents

Abstract

The present invention relates to a deep learning system for detecting disaster risks in power outlets. More specifically, it detects the risk of disasters such as fires in advance in underground power outlets where power cables for distribution or transmission are laid underground, and in the event of a fire, it is possible to quickly find the ignition point, thereby reducing the total power consumption. It relates to a system that can prevent the spread of fire within a district. Since the present invention can detect risk points with a high possibility of disaster occurrence in power outlets in advance using deep learning, it is possible to prevent the occurrence of disasters in advance through intensive monitoring activities for the risk points, as well as to prevent actual fires. When a fire alarm occurs due to a fire, it is possible to quickly and accurately find the ignition point and extinguish the fire. Therefore, it is effective to quickly find the ignition point in the power outlet where the cable is laid over a long distance, and it is possible to quickly extinguish the fire in the early stage of the fire, thereby preventing the fire from spreading to the entire power outlet.

Description

Deep learning system for disaster detection in the underground power facility}

The present invention relates to a deep learning system for detecting disaster risks in power outlets. More specifically, it detects the risk of disasters such as fires in advance in underground power outlets where power cables for distribution or transmission are laid underground, and in the event of a fire, it is possible to quickly find the ignition point, thereby reducing the total power consumption. It relates to a system that can prevent the spread of fire within a district. Since the present invention can detect risk points with a high possibility of disaster occurrence in power outlets in advance using deep learning, it is possible to prevent the occurrence of disasters in advance through intensive monitoring activities for the risk points, as well as to prevent actual fires. When a fire alarm occurs due to a fire, it is possible to quickly and accurately find the ignition point and extinguish the fire. Therefore, it is effective to quickly find the ignition point in the power outlet where the cable is laid over a long distance, and it is possible to quickly extinguish the fire in the early stage of the fire, thereby preventing the fire from spreading to the entire power outlet.

Due to the overcrowding of cities, improvement of urban aesthetics, and increased interest in the living environment of urban residents, the undergrounding rate of various urban infrastructure facilities, including power and communication lines, gas supply lines, district heating pipelines, etc., is rapidly increasing. These underground infrastructure facilities are sometimes concentrated and installed in urban infrastructure such as underground power outlets, communication outlets, or utility outlets. Among them, the underground utility tunnel is an underground facility for integrating and installing underground infrastructure elements, and the underground power tunnel is a facility in which a plurality of power transmission or distribution cables are installed.

Electric power outlets and utility outlets are high-risk environments with risks such as high voltage and high heat. However, when utility outlets and power outlets are damaged, the infrastructure in the area is paralyzed, resulting in widespread inconvenience such as power outages, communication disruptions, and water outages. Therefore, utility tunnels and power outlets are designated as important national facilities, and fire prevention measures are established, such as using fireproof materials. However, since the inside of the utility tunnel is narrow and closed, when a fire breaks out, smoke fills up in an instant, making it difficult to extinguish.

When a problem occurs in a power outlet or utility tunnel, it is inevitable for workers to visit the site for measures such as suppression or resolution. There are practical difficulties in Therefore, a surveillance system capable of monitoring various risk factors must be provided.

Therefore, various monitoring equipment capable of detecting the occurrence of a disaster such as a fire and suppression facilities such as sprinklers are installed in utility outlets including power outlets. However, existing surveillance equipment can not only identify disasters only after they occur. Even in the event of a disaster, there is a problem in that it is difficult to find the exact location of the occurrence and suppress it. Accordingly, there is still a problem in that a patrol personnel or a person in charge of a task has to directly move to the site and visually check it. However, since the power outlet is narrow and long, it takes a considerable amount of time as well as limitations for a person to visually check each item while walking. It is practically impossible because there is

Registered Patent Publication No. 10-1259478 (2013.05.06) Registered Patent Publication No. 10-1838193 (2018.03.13) Registered Patent Publication No. 10-1765036 (2017.08.04.) Registered Patent Publication No. 10-2354704 (2022.01.19.)

Chuxu Zhang et al., “A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data,” arXiv:1811.08055v1 (2018.11.20.)

The present invention, which was invented to solve the above problems, collects temperature distribution data in a power outlet and learns it as a deep learning model, and the learned deep learning model is used to predict abnormal situations and disasters such as fire Its purpose is to provide an information system that can detect and inform the occurrence location.

Another object of the present invention is, when the control server receives a fire detection signal or a fire alarm signal from an automatic fire detection system or a fire alarm system in a power outlet, the moving device moves inside the power outlet and determines the exact location of the ignition point. Its purpose is to provide an information system capable of finding and transmitting the location to the control server when an ignition point is found.

In addition, when the mobile device finds the ignition point and informs the control server, the control server informs the fire suppression equipment in the power outlet to provide an information system that enables quick fire suppression at the ignition point. The purpose.

Another object of the present invention is to prevent flood damage to electric power facilities located in places other than the ignition point by enabling fire suppression facilities in power outlets to intensively extinguish fires at the ignition point when the location of the ignition point is found. It aims to provide an information system that can

Another object of the present invention is to precisely detect the vicinity of the location where the abnormal situation occurs after the mobile device quickly moves to the location where the abnormal situation occurs when the deep learning model receives a fire detection signal after detecting an abnormal situation. The purpose of the present invention is to provide an information system capable of quickly and accurately locating an ignition point and extinguishing a fire therefor by finding the exact location of an ignition point.

Another object of the present invention is to accurately and precisely measure the temperature of each position inside the power bulb while the moving device can move quickly along the power bulb when collecting the temperature distribution in the power bulb formed over a long distance. It aims to provide an information system that can

Another object of the present invention is information that can provide location information on an ignition point in three-dimensional coordinates within a power outlet so that it can be found and suppressed using accurate location information on the ignition point within the power outlet. It aims to provide a system.

Another object of the present invention is to provide an information system capable of early suppression of a fire in the early stages of a fire by allowing the mobile device to spray an extinguishing solution directly to the ignition point when the mobile device finds the exact location of the ignition point. intended to provide

Another object of the present invention is to provide an information system capable of finding the exact location of an abnormal situation occurrence location and ignition point without expensive video cameras and complex live-action image processing devices that take pictures of the inside of a power outlet. do.

Another object of the present invention is that the mobile device is normally in a charging mode, but is operated by a battery in case of fire, so that even when power supply is not available due to a fire, it is possible to find the location of the fire and extinguish the fire. It aims to provide an information system that can

The technical problems to be achieved by the present invention are not limited to the above-mentioned technical problems, and other technical problems not mentioned can be clearly understood by those skilled in the art from the description below. There will be.

The present invention for achieving this object,

A rail installed on the upper part along the length of the power outlet, a moving device that can move along the rail, a docking station that can charge the moving device by docking it at an end point of the rail, and parallel to the rail A disaster risk detection and suppression system in an electric power outlet including a temperature sensing strip attached to the sidewall of the power outlet with the same length as the rail and a control server connected to the mobile device through data communication, wherein the mobile device includes a control means and a communication means , a driving means, an infrared temperature sensor, a thermal imaging camera, a tilting means, and a moving distance measuring means, wherein the driving means moves the moving device on the rail under the control of the control means, and the infrared The temperature sensor measures the temperature value (t) of the surface of the temperature sensing strip and transmits it to the control means, and the thermal imaging camera mounted on the tilting means takes a thermal image according to the control of the control means, The tilting means adjusts the tilt of the thermal imaging camera by rotating up and down under the control of the control means and measures the adjusted tilting angle (θ), and the moving distance measuring means measures the docking The distance value (d) moved by the mobile device from the station is measured and transmitted to the control means, and the communication means receives data transmitted from the control server, transmits the data to the control means, and sends it from the control means. The data is transmitted so that the control server can receive it, and the control means operates the driving means so that the moving device can reciprocate between both ends of the rail at regular time intervals, and the moving device moves. while generating temperature distribution data including the temperature value (t) and the distance value (d) measured by the infrared temperature sensor and the moving distance measurement means, and transmitting the temperature distribution data to the control server, - the power from the control server When receiving a fire signal in the ward, direct the lens of the thermal imaging camera downward and zoom out to a wide angle, and then move the moving device to the end of the rail by controlling the driving means to obtain still thermal images at regular intervals. is photographed and transmitted to the control server together with the distance value (d) for the shooting place, - in case of receiving information on the location of suspected ignition from the control server, - the distance value (d) included in the location information of suspected ignition ), after moving the moving device back to the position corresponding to - while zooming in the lens of the thermal imaging camera, the center of the thermal image having the predetermined temperature or more is located at the center point of the screen transmitted by the thermal imaging camera. After controlling the tilting means and the driving means, position information on the ignition point is calculated using the tilting angle (θ) and the distance value (d), and then transmitted through the communication means, and the control The server, - detects whether and where an abnormal situation occurs in the power outlet while learning the temperature distribution data with an abnormal situation detection model, which is a deep learning model, and when the abnormal situation is detected, the manager of the control server identifies - When a fire detection signal or a fire alarm signal is received from an automatic fire detection system installed to detect whether a fire has occurred in the power outlet The fire outbreak signal is generated and transmitted to the mobile device, and - when the still thermal image and the distance value (d) for the shooting location are transmitted from the mobile device after the fire outbreak signal is transmitted, the received stationary signal is transmitted. After creating and analyzing the temperature distribution histogram for each of the thermal images, the distance value (d) to the location where the still thermal image with the largest number of pixels having a certain temperature or higher in the histogram was taken is used as the location information of the suspected ignition. Transmitted to a mobile device, - In the case of receiving location information on the ignition point after transmitting the fire signal, fire suppression is performed at the ignition point using the fire suppression equipment in the power outlet, or the control It is preferable to use a deep learning system for detecting disaster risk in electric power outlets, characterized in that the location information on the ignition point and the still thermal image are displayed so that the manager of the server can identify them.

In addition to the above features, the control server includes the occurrence location of the abnormal situation in the fire occurrence signal when receiving the fire detection signal or the fire alarm signal after the abnormal situation is detected. When the fire outbreak signal received from the control server includes the occurrence location of the abnormal situation, the control unit moves the moving device to the occurrence location of the abnormal situation at the first speed. After, - reciprocating a certain distance at the second speed around the location where the abnormal situation occurs, - stopping the movement of the moving device at the position where the temperature value t measured by the infrared temperature sensor is the highest After the lens of the thermal imaging camera is zoomed out and then zoomed in, the tilting means and the driving means are controlled so that the center of the thermal image having a certain temperature or higher is positioned at the center point of the screen transmitted by the thermal imaging camera. However, - The first speed is characterized in that the second speed is faster, a deep learning system for detecting disaster risk in an electric power outlet

In addition to the above features, the present invention also provides the temperature sensing strip, - a heat insulation layer attached to a surface in contact with the side wall of the power outlet, - a metal strip repeatedly attached to the heat insulation layer at a predetermined length, and - between the metal strips. It is also preferable to have a deep learning system for detecting disaster risk in electric power outlets, characterized in that it includes partition walls that insulate the metal strips from each other.

In addition to the above-described features, the present invention is also a three-dimensional solid coordinate (x, y, z) for the ignition point, - the x value is the distance value (d), -y value is the distance from the bottom surface of the power outlet directly below the rail to the center of the cable hanger installed on the side wall of the power outlet, and when the cable hanger is located on both sides of the power outlet, one side is negative and the other side is a positive number, and -z value is the height from the bottom of the power outlet calculated by the formula below, characterized in that, it is also preferable to use a deep learning system for detecting disaster risk in the power outlet.

z = h - h / tan θ (h: height of rail, θ: tilting angle)

In addition to the above features, the present invention also provides a storage tank for storing the fire extinguishing solution, a spray nozzle capable of spraying the fire extinguishing solution, an electronic valve for opening and closing the spray nozzle, and a spray angle of the spray nozzle. Further comprising an angle adjusting means for adjusting the, wherein the control means, when calculating the positional information on the ignition point, the electronic valve and It is also preferable to use a deep learning system for detecting disaster risk in a power outlet, characterized in that it controls the angle adjustment means.

)를 포함하는 다음과 같은 수학식으로 표현되며,In addition to the above-described features, the present invention also learns the temperature distribution data with a deep learning model of a convolutional autoencoder technique, estimates a threshold, and then estimates a threshold, It is also preferable to use a deep learning system for detecting a disaster risk in an electric power outlet, characterized in that the abnormal situation is detected from the temperature distribution data using the threshold value, and in addition, the autoencoder, the temperature distribution data The scaled data vector (

) is expressed by the following equation including,

: 스케일된 데이터,

: 재생성된 데이터,

: 잠재공간(latent space),

: 인코더 함수,

: 디코더 함수-

: scaled data,

: regenerated data,

: latent space,

: encoder function,

: decoder function

는

일 때 다음과 같으며,The algorithm used for the encoder and decoder is composed of layer convolution 1D, which is a neural network structure, and neurons, and the encoder function

Is

is as follows when

- kernel: size of filters of convolution, stride: size of movement, filters: number of output filters

는 다음과 같으며,decoder function

is as follows,

와

의 차이를 줄이기 위한 손실함수 MSE는 다음과 같은 것을 특징으로 하는, 전력구 내 재난위험을 탐지하는 딥러닝 시스템으로 하는 것도 바람직하다.in the process of learning

and

The loss function MSE to reduce the difference in is also preferably a deep learning system for detecting disaster risk in power outlets, characterized by the following.

와 인풋으로 사용한

와의 차이를 구하여 상기 임계치를 다음과 같이 추정하는 것을 특징으로 하는, 전력구 내 재난위험을 탐지하는 딥러닝 시스템으로 하는 것도 바람직하다.In addition to the above-described features, the present invention also provides that the threshold estimate is the final value derived after learning for a certain epoch.

and used as input

It is also preferable to use a deep learning system for detecting disaster risk in power outlets, characterized in that the threshold is estimated as follows by obtaining the difference between

As described above, the moving device included in the present invention includes a control means, a driving means, a communication means, and an infrared temperature sensor, and the control means allows the moving device to reciprocate between both ends of the rail at regular time intervals. While the driving means is operated, the temperature distribution data including the temperature value (t) and the distance value (d) measured by the infrared temperature sensor and the moving distance measuring means are generated and transmitted to the control server while the moving device is moving, The control server learns the temperature distribution data with the abnormal situation detection model, which is a deep learning model, and detects whether and where an abnormal situation occurs in the power outlet. Since disaster risk information including the location of occurrence is generated and displayed, it is possible to detect and notify abnormal situations where disasters such as fires are expected to occur within the power outlet and the location of the occurrence can be detected and reported to prevent disasters in the power outlet in advance. It works.

In addition, when receiving a fire outbreak signal from the control server, the moving device included in the present invention directs the lens of the thermal imaging camera downward, zooms out to a wide angle, and then controls the driving means to move the moving device to the opposite end of the rail while moving the moving device to the opposite end of the rail. Still thermal images are taken at distance intervals and transmitted to the control server along with the distance value (d). The control server creates and analyzes a temperature distribution histogram for each of the received still thermal images, and then selects a certain temperature or higher from the histogram. The location where the still thermal image with the largest number of pixels is captured is determined as the position suspected of ignition and transmitted to the moving device. The moving device moves the moving device back to the position suspected of ignition. After controlling the tilting means and the driving means so that the center of the thermal image having a certain temperature or higher is located at the center point of the screen transmitted by , the ignition point is calculated using the tilting angle (θ) and the distance value (d). It has the effect of quickly finding the exact location of the ignition point.

In addition, when the control means of the mobile device included in the present invention finds an ignition point, it can transmit the location information to the control server, and when the control server receives the location information on the ignition point, the fire suppression equipment in the power outlet Since it is possible to extinguish the fire at the point of ignition by using, there is an effect of being able to extinguish the fire early on the exact location when a fire alarm occurs.

In addition to this, since the location information on the ignition point is displayed so that the manager of the control server can identify it when the location information on the ignition point is received, the manager can accurately identify the fire occurrence point, and accordingly, necessary There is an effect that can be performed accurately.

In addition, the control server included in the present invention, when receiving a fire detection signal or a fire alarm signal after an abnormal situation is detected, transmits the fire occurrence signal including the location of the abnormal situation to the control unit, and the control unit transmits the fire occurrence signal to the control unit. If the fire signal received from the location where the abnormal situation occurred is included, the moving device is moved at the first speed (fast speed) to the location where the abnormal situation occurred, and then the second speed (a certain distance around the location where the abnormal situation occurred) Since the movement of the moving device is stopped at the position where the temperature value (t) is highest while moving back and forth at a slow speed), the moving device can quickly move to the location where the abnormal situation occurs, and once it arrives has the effect of finding the exact location of the ignition point by precisely detecting the vicinity of the location where the abnormal situation occurred.

In addition, the present invention has a temperature sensing strip attached to the sidewall of the power outlet parallel to the rail along the length direction of the power outlet, and the temperature sensing strip is repeatedly attached to the heat insulation layer attached to the surface in contact with the sidewall of the power outlet and on the heat insulation layer at a certain length. Since it has a feature of including a metal strip and a barrier rib that insulates each other between the metal strips, the temperature of the sidewall of the power outlet, which is not sensitive to temperature change, is not transmitted to the metal strip due to the insulation layer, and the metal strip is thin. Since it has a thickness, it responds sensitively to the change in ambient temperature, and since heat is not transferred to each other due to the barrier rib with neighboring metal strips, the metal strip of the temperature sensing strip can most accurately reflect the temperature of the corresponding point. In addition, since the surface temperature of the temperature sensing strip is measured with an infrared temperature sensor, the temperature can be quickly measured. Therefore, even if the mobile device moves and measures the temperature distribution within the power outlet, the temperature distribution can be quickly and accurately measured.

In addition, since the location information on the ignition point in the present invention can be represented by three-dimensional coordinates (x, y, z) for the power sphere, when finding the ignition point in the power sphere, find it using accurate location information It has a suppressive effect.

In addition, the moving device included in the present invention includes a storage tank for storing the fire extinguishing solution, a spray nozzle capable of spraying the fire extinguishing solution, an electronic valve for opening and closing the spray nozzle, and an angle adjusting means for adjusting the spray angle of the spray nozzle. In addition, the control means also has a feature of controlling the electronic valve and the angle adjusting means so that the fire extinguishing solution can be sprayed to the ignition point through the spray nozzle when the location information on the ignition point is calculated. When the exact location of the point is found, the fire extinguishing solution can be sprayed directly from the moving device to the ignition point, which has the effect of early suppression of the fire at the beginning of the fire.

In the present invention, the mobile device sends the accurate location information of the ignition point found while moving to the fire suppression device to extinguish the fire at the ignition point, so that only the injection means located at the ignition point is operated, and the other injection means are not operated. Since this can be done, there is an effect of minimizing damage due to the operation of the suppression device for facilities or equipment located in a location other than the ignition point among other power facilities or equipment in the power outlet.

The present invention also includes a battery in the mobile device, a charger in the docking station, the charger charges the battery when the mobile device is docked in the docking station, and the mobile device is operated by the battery. Therefore, there is an effect that can provide an automatic fire suppression system that can operate normally even when electricity supply is not smooth due to a fire.

1 shows an underground distribution cable or a power outlet in which an underground transmission cable is laid.
2 shows a configuration diagram of a deep learning system for detecting disaster risk in a power outlet according to the present invention.
3 is a perspective view of a state in which the moving device included in the present invention is mounted on a rail.
4 is a front view of a front view of a moving device mounted inside a power outlet.
5 is a side view of a state in which the moving device is mounted inside the power outlet as viewed from the side.
6 is a top plan view of a state in which the mobile device is mounted inside the power outlet.
7 illustrates a concept in which a mobile device measures a temperature distribution inside a power outlet.
8 illustrates a concept in which a deep learning model detects an anomaly using temperature distribution data.
9 is an example of a histogram of a temperature distribution of a thermal image captured while the mobile device is moving.
10 is a diagram for explaining a method of finding an ignition point using a thermal imaging camera.
11 illustrates a concept in which the control means precisely detects an ignition point by controlling the speed of the moving device.
12 illustrates a concept of measuring the surface temperature of a temperature sensing strip with an infrared temperature sensor.
13 illustrates a detailed structure of a temperature sensing strip.
14 illustrates a method of obtaining a z value among three-dimensional coordinates using a tilting angle of a thermal imaging camera.
15 illustrates a method of obtaining an x value among three-dimensional coordinates using a moving distance of a mobile device.
16 shows an embodiment including a spray nozzle and an angle adjusting means in a moving device.
17 illustrates the concept of adjusting the spray angle of the spray nozzle with the angle adjusting means of the moving device.

Hereinafter, the present invention will be described in detail so that the above-described objects and characteristics become clear, and accordingly, those skilled in the art to which the present invention belongs will be able to easily implement the technical spirit of the present invention. In addition, in describing the present invention, among the known technologies related to the present invention, the detailed description is given when it is determined that the detailed description of the known technology may unnecessarily obscure the subject matter of the present invention. to omit

In addition, the terms used in the present invention have been selected from general terms that are currently widely used as much as possible, but in certain cases, there are terms arbitrarily selected by the applicant. It is intended to clarify that the present invention should be understood as the meaning of the term, not the name of. Terms used in the description of the embodiments are only used to describe specific embodiments, and are not intended to limit the embodiments. Singular expressions include plural expressions unless the context clearly dictates otherwise.

Embodiments may be changed in various forms and may have various additional embodiments. Here, specific embodiments are shown in the drawings and related detailed descriptions are described. However, this is not intended to limit the embodiments to a specific form, and it should be understood to include all changes, equivalents, or substitutes included in the spirit and technical scope of the embodiments.

Expressions such as “first”, “second”, “first” or “second” in the description of various embodiments may modify various components of the embodiments, but do not limit the corresponding components. For example, the above expressions do not limit the order and/or importance of corresponding components. The above expressions may be used to distinguish one component from another.

Hereinafter, the present invention will be described with reference to the accompanying drawings. 1 shows a general power outlet 10 in which cables for underground distribution or cables for underground transmission are installed. The power outlet 10 is shaped like a tunnel formed over a long distance in an underground space in order to lay an underground distribution line or a power cable 16 constituting an underground transmission line, and the power outlet 10 A cable hanger 15 having one or more layers is installed on the side wall to stably lay the power cable 16, and the cable hanger 15 is usually installed on both sides. The power cable 16 is placed on the cable hanger 15. Since transmission power or distribution power is supplied in three phases, all power cables 16 constituting three phases are stacked on one cable hanger 15. A voltage higher than the special high voltage or super special high voltage is applied to all of the power cables 16, and a large load current flows.

In this way, the power cable 16 having a voltage higher than the special high voltage or extra special high voltage is closely stacked in the power outlet 10, and a large current flows inside the power cable 16, so that the power cable 16 has a load. Heat generation by electric current occurs, and various phenomena caused by the application of special high pressure or super special high pressure cause heat generation and deterioration of the outer skin, so there is a risk of fire. In particular, this risk is higher due to poor contact or poor connection in the cable connection part or cable head. On the other hand, since the power outlet 10 is in the form of a narrow and very long tunnel, not only is it difficult to inspect or monitor in normal times, but also smoke spreads to the entire power outlet in an instant and the fire spreads in an instant when a fire occurs. Therefore, it is important to respond after a disaster occurs, but it is also necessary to prepare a system that can predict the signs of a disaster in advance and, in the event of a disaster, quickly and accurately detect and suppress it in the early stages. , the present invention is for this purpose.

2 shows a configuration diagram of a deep learning system for detecting disaster risk in a power outlet according to the present invention, and FIG. 3 shows a mobile device 200 included in the present invention mounted on a rail 100. A perspective view is shown. 4 is a front view of the moving device 200 mounted inside the power outlet 10, viewed from the front, and FIG. 5 is a side view showing the moving device 200 mounted inside the power outlet 10. 6 is a top plan view of the moving device 200 mounted inside the power outlet 10, as viewed from above.

As shown in FIGS. 2 to 6, the present invention includes a rail 100 installed on the upper part along the longitudinal direction of the power outlet 10, a moving device 200 capable of moving along the rail 100, the A docking station 110 capable of docking and charging the moving device 200 at one or both ends of the rail 100, attached to the sidewall of the power outlet parallel to the rail and the same length as the rail It is preferable to use a system including a temperature sensing strip 150 and a control server 300 connected to the mobile device 200 through data communication.

The mobile device 200 is preferably connected to the control server 300 through a wireless Internet or IoT communication network to communicate with it. However, the mobile device 200 communicates with the docking station 110 through wifi, wireless Internet or Internet of Things communication network, and the docking station 110 and the control server 300 are connected through the Internet or the like. It is also desirable to

After the mobile device 200 is equipped with a battery (not shown) and the docking station 110 includes a charging device, the mobile device 200 is docked with the docking station 110 At this time, it is more preferable to connect through the charging socket 280 to charge the battery, and to operate the mobile device 200 by the battery.

On the other hand, the control server 300 is connected to the automatic fire detection system 400 capable of generating a fire detection signal or a fire alarm signal when a fire occurs in the power outlet, so that when a fire occurs in the power outlet, the It is preferable to receive a fire detection signal or a fire alarm signal from the automatic fire detection system 400 . In addition, the control server 300 is connected to the fire suppression equipment 500 capable of extinguishing a fire in the event of a fire in the power outlet, thereby providing location information about an ignition point to the fire suppression equipment 500. Therefore, it is desirable to be able to extinguish the fire at the exact ignition point.

The automatic fire detection system 400 may transmit an automatic fire detection means 410 such as a fire detector, an alarm means 420 such as a fire alarm button, and the fire detection signal and the fire alarm signal to the control server 300. It is desirable to include a communication means 430 there. In addition, the fire suppression equipment 500 includes a communication means 510 capable of communicating with the control server 300 or the automatic fire detection means 410, and a spraying means capable of spraying fire extinguishing water or liquid, such as a sprinkler. 530 and a control means 520 capable of controlling the injection means 530 according to a signal sent from the control server 300 or the automatic fire detection means 410.

As shown in FIG. 2, the moving device 200 included in the present invention is a rail installed on the upper portion along the longitudinal direction of the power outlet 10 so as to be movable along the longitudinal direction of the power outlet 10. It is preferable to have it mounted on (100). In addition, the moving device 200 includes a control means 210, a communication means 230, a driving means 240, an infrared temperature sensor 220, a thermal imaging camera 250, a tilting means 251, a movement It is preferable to include a distance measuring means 260 and a battery (not shown).

As shown in FIG. 2, the cross section of the rail 100 is shaped like a ball, so that the moving device 200 is coupled to the bottom of the ball. The driving means 240 is configured to move the moving device 200 on the rail 100 according to the control of the control means 210 . To this end, the drive means 240 includes a drive motor 241 that generates rotational force, a drive roller 243 rotated by the drive motor 241, and auxiliary rollers 242 located on both sides of the drive roller 243. ), and the driving roller 243 and the auxiliary roller 242 are preferably rolled on top of a portion protruding on both sides under the ball shape of the rail 100.

In addition, the infrared temperature sensor 220 measures the temperature value t of the surface of the temperature sensing strip 150 while the moving device 200 moves and transmits the measured temperature value t to the control means 210, wherein the movement While the device 200 moves along the rail 100, it is preferable to measure a change in the temperature value t according to a change in position. Therefore, the infrared temperature sensor 220 continuously outputs the temperature value t measured at the current position, and the control unit 210 continuously outputs the temperature value t output by the infrared temperature sensor 220. It is desirable to receive, store and utilize.

The thermal imaging camera 250 is mounted on the tilting unit 251 so as to be rotatable, and captures a thermal image under the control of the control unit 210 and transmits it to the control unit 210. It is desirable to do In addition, the tilting means 251 adjusts the tilt of the thermal imaging camera 250 by rotating up and down under the control of the control means 210, and measures the adjusted taleting angle θ to obtain the measured taleting angle ( θ) is preferably provided to the control means 210. Therefore, while the moving device 200 moves along the rail 100, the thermal imaging camera 250 rotates up and down under the control of the control means 210 to generate heat by the tilting means 251. It is preferable that the image capture angle is adjusted and the control unit 210 continuously outputs thermal image data input to its imaging tube while zooming in or zooming out according to control. Also, it is preferable that the control unit 210 continuously receives, stores, and utilizes the thermal image output from the thermal image camera 250 .

It is preferable that the moving distance measurement means 260 measures the distance value d that the mobile device 200 has moved from the docking station 110 and transmits it to the control means 210 . Therefore, the moving distance measuring means 260 counts the operating time or number of rotations of the driving motor 241 or the number of rotations of the driving roller 243 or the auxiliary roller 242, etc. While measuring the distance value (d) moved from the docking station 110, while the mobile device 200 leaves the docking station 110 and moves, the distance value (d) moved is continuously measured and output. It is preferable that the control unit 210 continuously receives and stores the distance value d outputted by the movement distance measuring unit 260 so as to be used.

The communication means 230 receives data transmitted from the control server 300 and transmits it to the control means 210, and the control server 300 receives the data sent from the control means 210. It is desirable to send it so that it can be done. Therefore, the communication means 230 is preferably directly connected to the control server 300 through a wireless Internet or Internet of Things communication network to communicate. However, the control means by allowing the communication means 230 to communicate with the docking station 110 through wifi, wireless Internet or Internet of Things communication network, etc., and connecting the docking station 110 to the control server 300. When 210 is connected to the docking station 110 through the communication means 230, the docking station 110 serves as a router to control the control server 300 through wired Internet or a wireless Internet network such as LTE or 5G. ) is also desirable.

Meanwhile, the control means 210 operates the driving means 240 so that the moving device 200 can reciprocate between both ends of the rail 100 at regular time intervals, and the moving device 200 While moving, the control server 300 generates temperature distribution data including a temperature value (t) measured by the infrared temperature sensor 220 and a distance value (d) measured by the moving distance measuring means 260. It is preferable to send it to That is, while the moving device 200 reciprocates between both ends of the rail 100, the control means 210 obtains the temperature value t according to the distance value d, Using the temperature value t according to the distance value d, as shown in FIG. 7 , temperature distribution data on both sides of the sidewall 13 of the power outlet 10 is generated.

At this time, the predetermined time is a time interval for checking the possibility of a disaster occurring in the power outlet 10, and is to be determined according to the situation of the power outlet 10. For example, in a power district 10 that needs to closely check the possibility of a disaster, it may be performed at dense intervals such as 1 hour intervals or 2 hour intervals, and in other power districts 10, 24 hour intervals or 3 day intervals It may be done at relatively distant intervals, such as And it will be possible to check it from time to time according to the request of the management agency. It is more preferable to set the predetermined time in consideration of the charging capacity of the moving device 200.

The temperature distribution data is data for the control server 300 to determine the possibility or danger of a disaster occurring in the power outlet 10, and the control server 300 uses the temperature distribution data as a deep learning model to The occurrence of an abnormal situation in the power outlet 10 is detected. To this end, the control server 300 detects whether and where an abnormal situation occurs in the power outlet 10 while learning the temperature distribution data with an abnormal situation detection model, which is a deep learning model, and detects the abnormal situation. In this case, it is preferable to generate and display disaster risk information including the occurrence location of the abnormal situation so that the administrator of the control server 300 can identify it.

That is, as shown in FIG. 8, the temperature distribution data (h1 to h6) sent at regular time intervals are learned with a deep learning learning model, and an abnormal situation is detected in the temperature distribution data according to the learning result. . For example, if it is determined that the part marked with a red ellipse in h6 of FIG. 8 shows a different temperature distribution than usual, the distance value (d1 to d2) of the part marked with a red ellipse is judged as an abnormal situation. It can be done. Here, the comparison target 'normal time' will be able to determine the comparison target in consideration of various variables such as season, time, load current, surface temperature, underground temperature, and air temperature. To this end, the control server 300 provides weather information, temperature It is more desirable to have means to receive data of various variables such as information, season information, and current information.

Even if the fire detection signal or fire alarm signal is not received from the automatic fire detection system, the present invention analyzes the temperature distribution data in the power outlet at ordinary times to find the temperature change trend, etc., and even includes the location where an abnormal situation occurs in the temperature distribution. Since disaster risk information is generated and displayed, it is possible to identify the point at which a disaster such as fire is predicted to occur before it occurs, and when disaster risk information is provided, managers, etc. Since it can be performed, there is an effect of fundamentally removing the factors of disaster occurrence in advance, such as blocking the possibility of a disaster occurring in the electric power outlet 10 in advance.

Meanwhile, FIG. 9 shows an example in which the temperature distribution of the thermal image taken while the moving device 200 is moving is shown as a histogram, and FIG. has been In the present invention, the control server 300 transmits a fire signal to the mobile device 200 when a fire occurs in the power outlet 10, and the control means included in the mobile device 200 ( 210) uses the thermal imaging camera 250, the tilting means 251, and the driving means 240 when receiving a fire occurrence signal in the power outlet 10 from the control server 300. When a still thermal image of the entire section within the power outlet 10 is photographed at a wide angle and sent to the control server 300 together with the shooting location, the control server 300 analyzes the still thermal image in a histogram to determine the occurrence of a fire. The photographed location is found and the location is sent to the moving device 200, and the moving device 200 uses the thermal imaging camera 250, the tilting means 251, and the driving means 240 to accurately After finding the ignition point, it is preferable to provide the location information to the control server 300 .

To this end, the control server 300 generates the fire occurrence signal when receiving a fire detection signal or a fire alarm signal from the automatic fire detection system 400 installed to detect whether a fire has occurred in the power outlet 10. When the mobile device 200 receives a fire occurrence signal in the power outlet 10 from the control server 300, the thermal imaging camera 250 After directing the lens of ) downward and zooming out to a wide angle, the driving means 240 is controlled to move the moving device 200 to the opposite end of the docking station 110 along the rail 100 while moving a certain distance. It is preferable to take a still thermal image at intervals and transmit it to the control server 300 together with the distance value d for the location.

Here, directing the lens of the thermal imaging camera 250 downward and zooming out to a wide angle is the power cable 16 installed on both sides of the power outlet 10 and the cable hangers 15 supporting them. It is for taking pictures at once, and taking a thermal image as a still image instead of a moving image is for the accuracy and speed of histogram analysis. In the case of shooting a thermal image as a moving image, it is not only time consuming to analyze the histogram for all frames (20 to 30 frames per second) that form the moving image, but it is also difficult to find the location of the shooting point for each frame of the moving image. On the other hand, as in the present invention, when a still image is taken without missing for each section of the electric power outlet 10, each of them is histogram-analyzed, and the shooting position is found, the histogram Analyzes can be performed quickly and accurately.

Meanwhile, the temperature distribution histogram is obtained by rearranging the pixels constituting the still thermal image according to the temperature after counting the number of pixels according to the temperature value. Therefore, when the control server 300 receives the distance value d for the still thermal image and the photographing location from the mobile device 200, as shown in FIG. 9, for each of the received still thermal images After making and analyzing the temperature distribution histogram, a still thermal image with the most pixels having a certain temperature or higher is selected from the histogram, and the distance value (position d4 in FIG. Preferably, it is transmitted to the mobile device 200 as information.

Then, when the control unit 210 of the mobile device 200 receives the suspected ignition location information from the control server 300, the location corresponding to the distance value d included in the location information of the suspected ignition location is displayed. After moving the moving device back, while zooming in the lens of the thermal imaging camera, the tilting means 251 and After controlling the driving means 240, calculating the location information on the ignition point using the tilting angle θ and the distance value d, and then transmitting it through the communication means 230 desirable.

At this time, as shown in FIG. 9(a), the control means 210 controls the tilting means 251 and the driving means 240 to repeat tilt adjustment and reciprocating movement so that the thermal imaging camera 250 transmits After the center of the thermal image having a certain temperature or higher is located at the center point of the screen, position information on the ignition point is calculated using the tilting angle (θ) and the distance value (d), and the communication means ( 230) is preferred. In addition, in this process, as shown in FIG. 9 (b), the tilting means 251 and the driving means 240 are controlled to find the positions of both ends (d3, d4) and the upper and lower positions of the ignition point, It is also possible to calculate and provide the width (b = d4 - d3) and height (a = c2 - c2) of the ignition point (height is calculated using a tilt angle change).

On the other hand, when the control server 300 receives the location information on the ignition point after transmitting the fire occurrence signal, the fire suppression for the ignition point is suppressed by using the fire suppression equipment in the power outlet 10. It is desirable to perform In addition, it is preferable to display location information about the ignition point and the still thermal image so that the administrator of the control server 300 can identify them.

In the present invention, when the control means 210 included in the mobile device 200 receives a fire detection signal or a fire alarm signal through the communication means 230, the driving device 240 is operated to When the moving device 200 moves along the rail 100 and takes a still thermal image and transmits it to the control server 300, the control server 300 finds a suspected ignition point using the still thermal image and Informing the mobile device 200, the mobile device 200 moves to the ignition-suspicious point and aligns the center of the thermal imaging camera 250 with the center of the thermal image while reciprocating around the ignition-suspicious point. Since the tilting angle (θ) of the image camera 250 is measured, and location information on the ignition point is calculated using the tilting angle (θ) and the moving distance (d), the exact location of the ignition point can be quickly and accurately determined in the event of a fire. There is an effect that can be found.

When aligning the center of the thermal image camera 250 with the center of the thermal image, it is also preferable to use a technology for automatically identifying and tracking the shape of a target such as an object or a person's face in a camera. That is, when various shapes of flames or heating elements generated by fire are stored, and the flame shape or heating element shape is captured on the screen of the thermal imaging camera 250, the control unit 210 controls the shape of the flame. The driving means 240 and the tilting means 251 are controlled so that the center comes to the center point of the thermal image screen. However, it is also possible to use the temperature value for each pixel of the thermal image screen. For example, pixels having the highest temperature are located at the center of the thermal image screen. In addition to this, the center of the thermal image camera 250 may be aligned with the center of the thermal image using various known techniques.

In addition, when the ignition point is found, the location information can be transmitted to the control server 300, and the control server 300 can suppress the fire at the ignition point using the fire suppression equipment in the power outlet 10. Therefore, when a fire alarm occurs, there is an effect of being able to quickly extinguish the fire at the exact location of occurrence. In addition, since the location information on the ignition point is displayed so that the manager of the control server can identify it when the location information on the ignition point is received, the manager can accurately identify the fire occurrence point and take necessary measures accordingly. There is an effect that can be performed accurately.

On the other hand, in the present invention, when a fire occurs after the abnormal situation is detected by the deep learning model, the entire section of the power outlet 10 is not searched, and the abnormal situation is quickly moved to the point where it is detected. It also includes a method for finding the ignition point more quickly and accurately in the event of a fire by closely inspecting the surroundings around the point where the abnormal situation is detected to find the ignition point. 11 illustrates a concept in which the control unit 210 precisely detects an ignition point by controlling the moving speed of the moving device 200 in the present invention.

That is, when the control server 300 receives the fire detection signal or the fire alarm signal after the abnormal situation is detected, the mobile device 200 includes the occurrence location of the abnormal situation in the fire occurrence signal. ) to the control means 210, and the control means 210, when the fire occurrence signal received from the control server 300 includes the occurrence location of the abnormal situation, as shown in FIG. After moving the moving device at a first speed, which is a high speed, to the location where the abnormal situation occurs, it reciprocates a certain distance around the location where the abnormal situation occurs at a second speed (low speed), and then the infrared temperature sensor After stopping the movement of the moving device at the position where the temperature value t measured by 220 is the highest, while zooming out and then zooming in the lens of the thermal imaging camera 250, the thermal imaging camera 250 It is preferable to control the tilting means 251 and the driving means 240 so that the center of the thermal image having the predetermined temperature or higher is positioned at the center point of the screen to be transmitted. Here, the first speed is preferably higher than the second speed. And, if the ignition point is not found at the point where the abnormal situation is detected, it is more desirable to search the entire section by histogram analysis of the still thermal image as described above.

Since the present invention has such a configuration, the mobile device 200 can quickly move to the location where the abnormal situation occurs when a fire signal is received, and once it arrives, it precisely detects the vicinity of the location where the abnormal situation occurs to determine the ignition point. It has the effect of finding the exact location of the .

Meanwhile, in the present invention, while the moving device 200 is moving, the infrared temperature sensor 220 measures the temperature value t inside the power outlet 10 and provides it to the control means 210. Have. In the present invention, an infrared temperature sensor having a fast reaction speed is used to measure the temperature inside the power bulb 10 while the moving device 200 moves. This is because it is difficult to accurately measure the temperature of the current location where the device 200 moves.

In general, since the infrared temperature sensor measures infrared rays emitted from the surface of an object, the infrared temperature sensor 220 included in the present invention is the 'current temperature' at the 'current location' of the mobile device 200. It is desirable to measure the surface temperature of 'an object capable of reflecting ' on the surface. In addition, since the temperature measurement in the present invention is to measure the temperature distribution along the longitudinal direction of the electric power bulb 10, the 'object capable of reflecting the current temperature on the surface' is continuously along the longitudinal direction of the electric power bulb 10 It is desirable to make it an object that exists, has the same shape and is made of the same material.

As such, objects made of the same shape and material that continuously exist in the power outlet 10 include the ceiling 11, the side wall 13 and the floor 12 of the power outlet 10 and the rail 100 there is However, since the ceiling 11, the side wall 13 or the floor 12 of the power outlet 10 are concrete structures, they are not sensitive to temperature changes inside the power outlet 10, and heat storage from the previous point to the time of measurement. Since the temperature of the ceiling 11, the side wall 13, or the floor 12 is not the current temperature of the location, but also a temperature value that reflects the energy supplied from the surroundings because it is connected to one structure, It does not accurately reflect the current temperature of the point to be measured. In addition, since the rail 100 is a metal component and is connected as a whole, it does not reflect only the temperature of the corresponding point due to fast heat conduction, but also reflects the temperature of the surrounding point. There is a problem that cannot be reflected.

Therefore, in the present invention, even if the moving device 200 continuously moves and measures the temperature in the power outlet 10, in order to quickly and accurately measure the temperature at the current point, the sidewall 13 of the power outlet 10 ) and has a configuration in which the temperature sensing strip 150 is attached to the infrared temperature sensor 220 and the surface temperature of the temperature sensing strip 150 is measured. 12 shows the concept of measuring the surface temperature of the temperature sensing strip 150 with the infrared temperature sensor 220, and FIG. 13 shows a detailed structure of the temperature sensing strip 150.

That is, as shown in FIG. 12, the present invention has a constant width on the sidewall 13 of the power outlet 10 and is attached parallel to the rail 100 along the longitudinal direction of the power outlet 10. It is to have a sensing strip 150. And, as shown in FIG. 13, the temperature sensing strip 150 is a strip-shaped heat insulating layer continuously attached to a surface in contact with the sidewall 13 of the power outlet 10 along the longitudinal direction of the power outlet 10. 151, and it is preferable to repeatedly attach a metal strip 152 having a certain length on the heat insulating layer 151 along the longitudinal direction of the power outlet 10. Further, it is more preferable to include barrier ribs 153 between the metal strips 152 to insulate the metal strips 152 from each other. Further, the metal strip 152 preferably has a thin thickness of at least 1 mm or less, and the length of one metal strip 152 is about 1 m. In this way, each of the metal strips 152 becomes a means for measuring the temperature at intervals of 1 m within the power outlet 10 . However, the thickness and length may be varied according to environmental conditions.

Since the present invention has such a feature, the heat or cold of the sidewall 13 of the power outlet 10, which does not reflect the current temperature of the power outlet, can be blocked due to the heat insulation layer 151, and the metal having a thin thickness The band 152 not only reacts sensitively to temperature changes and reflects the current temperature, but also does not reflect the temperature change at the next point because it is insulated from other adjacent metal bands 152 by the barrier rib 153. It will be able to reflect only the current temperature for the point where it is located. Therefore, when the surface temperature of the metal strip 152 is measured by the infrared temperature sensor 220, the accurate temperature of the measurement point can be quickly measured. Therefore, even if the moving device 200 measures the temperature distribution within the power outlet 10 while moving, the temperature distribution can be quickly and accurately measured.

On the other hand, in the present invention, when providing location information on an ignition point, it has a configuration that can be provided as three-dimensional coordinates (x, y, z) for the power sphere by using the tilting angle of the thermal imaging camera. Therefore, When finding an ignition point in a power tunnel, it has the effect of finding the exact location and suppressing it. 14 shows a method of obtaining a z value among three-dimensional coordinates using the tilting angle (θ) of the thermal imaging camera 250 in the present invention, and FIG. 15 shows the movement distance (d) of the moving device 200 A method of obtaining an x value among three-dimensional coordinates using is shown.

That is, in the present invention, as shown in FIGS. 14 and 15, the location information on the ignition point is a three-dimensional coordinate (x, y, z) for the power sphere, and the x value is the distance value (d) And, the y value is the distance from directly below the rail of the bottom surface of the power outlet to directly below the center of the cable hanger installed on the side wall of the power outlet, when the cable hanger is located on both sides of the power outlet, one side is negative and the other One side is a positive number, and the z value is preferably the height from the bottom of the power outlet calculated by the formula below.

z = h - h / tan θ (h: height of rail, θ: tilting angle)

In addition, in the present invention, when the mobile device 200 finds the exact location of the ignition point, the mobile device 200 can directly spray the fire extinguishing solution to the ignition point, thereby extinguishing the fire in the early stage of the fire. It also provides a configuration that allows you to do this. To this end, the moving device 200 included in the present invention includes a storage tank 290 for storing the fire extinguishing solution, a spray nozzle 270 capable of spraying the fire extinguishing solution, and electrons for opening and closing the spray nozzle 270. It is preferable to include a valve (not shown) and an angle adjusting means 271 for adjusting the spray angle of the spray nozzle 270.

16 shows an embodiment in which the moving device 200 includes the spray nozzle 270 and the angle adjusting unit 271, and FIG. 17 shows the spray nozzle 270 as the angle adjusting unit 271. ) It shows how to adjust the spray angle of. As shown in FIGS. 16 and 17, in the present invention, when the control means 210 included in the moving device 200 calculates the location information on the ignition point, the fire extinguishing solution is applied to the spray nozzle 270 ), it is preferable to control the solenoid valve and the angle adjusting means 271 so that the fuel can be injected at the ignition point.

On the other hand, as described above, the present invention provides a configuration for detecting an abnormal situation in the power outlet 10 using deep learning. To this end, the abnormal situation detection model is used. Hereinafter, the abnormal situation detection model will be described. Machine learning or deep learning, one of the technologies of artificial intelligence, enables computers to learn and make decisions on their own through past data. Basically, it learns the correct answer through given data in the past, and if similar data is given in the future, it will guess the answer. This method is called supervised learning, and deep learning is a similar logic structure that humans conclude, and has recently been in the limelight for its excellent performance by learning with a more complex method called artificial neural networks.

In deep learning, self-supervised learning is a method of learning in which the correct answer becomes input data. It is often used to detect outliers by extracting only important features of the data used as input data and making it similar to the input data. learning method used. On the other hand, AutoEncoder is a neural network that copies input data to output. It is a method of training a neural network by applying various constraints in this process. These constraints control the deep learning model to learn how to represent data more efficiently. Because of these characteristics, it can be used in anomaly detection, dimensionality reduction, or preprocessing.

An autoencoder is composed of an encoder and a decoder. The encoder converts input data into an internal representation, and the decoder reconstructs the internal representation into output data. In addition, the autoencoder is a representative semi-supervisor learning method, and its purpose is to learn the closest value of the identity function. That is, the purpose is to make the input value and the output value have the most similar form. Autoencoders learn how to represent data efficiently. The data of the autoencoder reconstructs the input data into compressed data of a lower dimension than the input. It is to find a certain structure and correlation between features in this data, and finally reconstruct the output data to be as similar to the input data as possible. An autoencoder can also be created by stacking several layers, and learning occurs for each layer. Models based on autoencoders are mainly used in various fields such as feature extraction, noise removal, and anomaly detection.

The abnormal situation detection model included in the present invention is a neural network that copies input data to output using an autoencoder. That is, a model that learns the temperature distribution data with a deep learning learning model of a convolutional autoencoder technique to estimate a threshold, and then uses the threshold to detect the abnormal situation from the temperature distribution data. do it with

In the case of learning, it is desirable to perform dimensionality reduction to extract only necessary characteristics, and to generate output data to have the same size as the input data to learn to obtain a loss. An auto encoder, which is a deep learning learning model, is divided into encoder and decoder parts, and can be expressed by the following equation including a scaling data vector.

는 스케일된 데이터이며,

는 재생성된 데이터이고,

는 잠재공간(latent space)이며,

와

는 인코더와 디코더의 역할을 하는 함수이다.here

is the scaled data,

is the regenerated data,

is the latent space,

and

is a function that serves as an encoder and a decoder.

의 일부분은 다음과 같은 수학식으로 나타낼 수 있다.Algorithms used in the encoder and decoder are composed of layers and neurons, which are neural network structures, and a layer called Convolution 1D is used in the present invention. The part corresponding to the encoder

A part of can be represented by the following equation.

일 때,

when,

는 다음과 같이 나타내는 것이 바람직하다.Here, kernel is the size of the convolution's filters, stride is the size of the movement, and filters are the number of output filters, reducing the size of the dimension. On the other hand, the function corresponding to the decoder

is preferably expressed as follows.

와

의 차이를 줄이기 위해 손실함수로는 MSE를 사용하며, 상기 MSE는 다음과 같이 하는 것이 바람직하다.The part of Encoder is calculated inversely, and the size of the dimension increases on the contrary. In the process of learning from the deep learning learning model

and

In order to reduce the difference in , MSE is used as the loss function, and the MSE is preferably as follows.

After performing the learning, it is preferable that the control server estimate a threshold for detecting an outlier. It is preferable to estimate the threshold value as follows by obtaining the difference between the final Y derived after learning for a certain epoch and X' used as an input.

After the estimation of the threshold value is performed, it is preferable to detect the anomaly value using the threshold value in order to detect the anomaly. Preferably, the outlier detection is performed by transmitting the data input in time series to the learned deep learning learning model and comparing the estimated threshold value and value as follows.

After performing the outlier detection, it is preferable to set the outlier detection result as the abnormal situation.

Although the present invention has been described with the various examples described above, the present invention is not necessarily limited to these examples, and may be variously modified and implemented without departing from the technical spirit of the present invention. Therefore, the examples disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these examples. The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the equivalent scope should be construed as being included in the scope of the present invention.

10 power district
11 ceiling 12 floor
13 side wall 15 cable hanger
16 power cable
100 rail
110 docking station
111 means of communication 112 means of charging
150 Temperature Sensing Band
151 insulation layer 152 metal strip
153 bulkhead
200 Mobility
210 control means 220 infrared temperature sensor
230 means of communication 240 means of drive
241 drive motor 242 auxiliary roller
243 drive roller
250 Thermal imaging camera 251 Tilting means
260 Moving distance measurement means 270 Spray nozzle
271 angle adjustment means 280 charging socket
290 storage tank
300 control server
400 Automatic fire detection system
410 Automatic fire detection means 420 Alarm means
430 means of communication
500 Fire suppression system
510 means of communication 520 means of control
530 dispensing means

Claims (8)

A rail installed on the upper part along the length of the power outlet, a moving device that can move along the rail, a docking station that can charge the moving device by docking it at an end point of the rail, and parallel to the rail A disaster risk detection system in a power outlet including a temperature sensing strip attached to the sidewall of the power outlet with the same length as the rail and a control server connected to the mobile device through data communication,
The moving device includes a control unit, a communication unit, a driving unit, an infrared temperature sensor, a thermal imaging camera, a tilting unit, and a moving distance measuring unit,
The driving means moves the moving device on the rail under the control of the control means,
The infrared temperature sensor measures the temperature value (t) of the surface of the temperature sensing strip and transmits it to the control means;
The thermal imaging camera mounted on the tilting means takes a thermal image under the control of the control means and transmits it to the control means;
The tilting means adjusts the tilt of the thermal imaging camera by rotating up and down under the control of the control means, and measures the adjusted tilting angle (θ) .
The moving distance measuring means measures a distance value (d) moved by the mobile device from the docking station and transmits it to the control means;
The communication means receives data transmitted from the control server, transmits it to the control means, and transmits the data sent from the control means so that the control server can receive it;
The control means,
- Operate the drive means so that the moving device can reciprocate between both ends of the rail at regular time intervals, and the temperature value measured by the infrared temperature sensor and the moving distance measuring means while the moving device is moving ( t) and temperature distribution data including the distance value (d) are generated and transmitted to the control server;
- When receiving a fire occurrence signal in the power outlet from the control server, after directing the lens of the thermal imaging camera downward and zooming out to a wide angle, the driving means is controlled to move the moving device to the opposite end of the docking station. While moving along the rail, still thermal images are taken at regular distance intervals and transmitted to the control server together with a distance value (d) for the shooting location,
- In case of receiving suspected ignition location information from the control server,
-- After moving the moving device back to the position corresponding to the distance value (d) included in the location information of suspected ignition,
-- While zooming in the lens of the thermal imaging camera, controlling the tilting means and the driving means so that the center of the thermal image having a certain temperature or higher is located at the center point of the screen transmitted by the thermal imaging camera,
-- Using the tilting angle (θ) and the distance value (d), location information on the ignition point is calculated and transmitted through the communication means;
The control server,
- While learning the temperature distribution data with an abnormal situation detection model, which is a deep learning model, it detects whether and where an abnormal situation occurs in the power outlet, and when the abnormal situation is detected, the manager of the control server can identify it Disaster risk information including the occurrence location of the abnormal situation is generated and displayed,
- When a fire detection signal or a fire alarm signal is received from an automatic fire detection system installed to detect whether a fire occurs in the power outlet, the fire occurrence signal is generated and transmitted to the mobile device,
- When the still thermal image and the distance value (d) for the shooting location are received from the mobile device, a temperature distribution histogram for each of the received still thermal images is created and analyzed, and then a certain temperature or higher is selected from the histogram. A distance value (d) to a location where a still thermal image having the largest number of pixels is captured is transmitted to the mobile device as the ignition suspicion location information;
- In the case of receiving location information on the ignition point, fire suppression is performed at the ignition point using the fire suppression equipment in the power outlet, or the location of the ignition point is identified so that the manager of the control server can identify it. Information and the still thermal image are expressed,
The abnormal situation detection model learns the temperature distribution data with a deep learning learning model of a convolutional autoencoder technique to estimate a threshold, and then uses the threshold to estimate the temperature distribution data from the temperature distribution data. detect anomalies,
The autoencoder is a data vector obtained by scaling the temperature distribution data (

) is expressed by the following equation including,

-

: scaled data,

: regenerated data,

: latent space,

: encoder function,

: decoder function
The algorithm used for the encoder and decoder is composed of layer convolution 1D, which is a neural network structure, and neurons, and the encoder function

Is

is as follows when

- kernel: size of filters of convolution, stride: size of movement, filters: number of output filters
decoder function

is as follows

in the process of learning

and

The loss function MSE for reducing the difference in

According to claim 1,
When the control server receives the fire detection signal or the fire alarm signal after the abnormal situation is detected, the control server transmits the fire occurrence signal including the occurrence location of the abnormal situation to the control means,
The control means, when the fire occurrence signal received from the control server includes the occurrence location of the abnormal situation,
- After moving the moving device at a first speed to the location where the abnormal situation occurs,
- While reciprocating a certain distance at the second speed centered on the location of the abnormal situation,
- After stopping the movement of the moving device at the position where the temperature value (t) measured by the infrared temperature sensor is the highest
- While zooming out and zooming in the lens of the thermal imaging camera, controlling the tilting means and the driving means so that the center of the thermal image having the predetermined temperature or higher is located at the center point of the screen transmitted by the thermal imaging camera,
-Deep learning system for detecting disaster risk in power outlets, characterized in that the first speed is faster than the second speed According to claim 1,
The temperature sensing belt,
- a heat insulating layer attached to a surface in contact with the side wall of the power outlet;
- A metal strip repeatedly attached with a certain length on the heat insulating layer, and
- Deep learning system for detecting disaster risk in electric power outlets, characterized in that it comprises a partition wall that insulates the metal strips from each other between the metal strips. According to claim 1,
The location information for the ignition point is a three-dimensional coordinate (x, y, z) for the power sphere,
- the x value is the distance value (d),
-y value is the distance from the bottom surface of the power outlet directly below the rail to the center of the cable hanger installed on the side wall of the power outlet, and when the cable hanger is located on both sides of the power outlet, one side is negative and the other side is a positive number,
- Deep learning system for detecting disaster risk in a power outlet, characterized in that the z value is the height from the bottom of the power outlet calculated by the following formula
z = h - h / tan θ (h: height of rail, θ: tilting angle) According to claim 1,
The moving device further includes a storage tank for storing the fire extinguishing solution, a spray nozzle capable of spraying the fire extinguishing solution, an electronic valve for opening and closing the spray nozzle, and an angle adjusting means for adjusting a spray angle of the spray nozzle,
The control means controls the solenoid valve and the angle adjusting means so that the fire extinguishing solution can be sprayed to the ignition point through the injection nozzle when the location information on the ignition point is calculated. , Deep Learning System to Detect Disaster Risk in Power Conduits delete delete delete

Images